NFD uses some Kubernetes custom resources.
NodeFeature is an NFD-specific custom resource for communicating node features and node labeling requests. The nfd-master pod watches for NodeFeature objects, labels nodes as specified and uses the listed features as input when evaluating NodeFeatureRules. NodeFeature objects can be used for implementing 3rd party extensions (see customization guide for more details).
apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeature
metadata:
labels:
nfd.node.kubernetes.io/node-name: node-1
name: node-1-vendor-features
spec:
features:
instances:
vendor.device:
elements:
- attributes:
model: "xpu-1"
memory: "4000"
type: "fast"
- attributes:
model: "xpu-2"
memory: "16000"
type: "slow"
labels:
vendor-xpu-present: "true"
NodeFeatureRule is an NFD-specific custom resource that is designed for rule-based custom labeling of nodes. NFD-Master watches for NodeFeatureRule objects in the cluster and labels nodes according to the rules within. Some use cases are e.g. application specific labeling in a specific environments or being distributed by hardware vendors to create specific labels for their devices.
apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
name: example-rule
spec:
rules:
- name: "example rule"
labels:
"example-custom-feature": "true"
# Label is created if all of the rules below match
matchFeatures:
# Match if "veth" kernel module is loaded
- feature: kernel.loadedmodule
matchExpressions:
veth: {op: Exists}
# Match if any PCI device with vendor 8086 exists in the system
- feature: pci.device
matchExpressions:
vendor: {op: In, value: ["8086"]}
See the Customization guide for full documentation of the NodeFeatureRule resource and its usage.
The deployment/nodefeaturerule/samples/
directory contains sample NodeFeatureRule objects that replicate the built-in default feature labels generated by NFD. The sample rules can be used as a base to customize NFD feature labels. To use them in place of the the NFD built-in labels, the corresponding feature source(s) of nfd-worker should be disabled with the core.labelSources
configuration option.
When run with NFD-Topology-Updater, NFD creates NodeResourceTopology objects corresponding to node resource hardware topology such as:
apiVersion: topology.node.k8s.io/v1alpha1
kind: NodeResourceTopology
metadata:
name: node1
topologyPolicies: ["SingleNUMANodeContainerLevel"]
zones:
- name: node-0
type: Node
resources:
- name: cpu
capacity: 20
allocatable: 16
available: 10
- name: vendor/nic1
capacity: 3
allocatable: 3
available: 3
- name: node-1
type: Node
resources:
- name: cpu
capacity: 30
allocatable: 30
available: 15
- name: vendor/nic2
capacity: 6
allocatable: 6
available: 6
- name: node-2
type: Node
resources:
- name: cpu
capacity: 30
allocatable: 30
available: 15
- name: vendor/nic1
capacity: 3
allocatable: 3
available: 3
The NodeResourceTopology objects created by NFD can be used to gain insight into the allocatable resources along with the granularity of those resources at a per-zone level (represented by node-0 and node-1 in the above example) or can be used by an external entity (e.g. topology-aware scheduler plugin) to take an action based on the gathered information.